Wear Condition Evaluation And Prediction Of Milling Cutter Based On Integrated Learning Algorithm

In CNC machining,the machining accuracy of parts and the machining efficiency of CNC machine tools are closely related to the wear condition of the tool.Therefore,in the milling process,intelligent detection of tool wear condition,timely detection and replacement of blunt milling cutters,are of great significance for improving the processing quality of the workpiece and ensuring the processing efficiency of the machine tool.In this paper,combined with milling cutter processing data and integrated learning algorithms,a data-driven milling cutter wear condition prediction model is established,which provides a basis for tool change decisions.The main content includes signal acquisition and data processing,feature selection and dimensionality reduction,and the establishment of a milling cutter wear condition prediction model.(1)Tool wear mechanism analysis and tool wear condition sensitive feature extraction.First,analyze the main wear forms and mechanisms of the tool.Later,the tool face wear VB value will be used as the milling tool wear standard.The trend changes of various sensor signals and wear values are analyzed,and the invalid value and abnormal value of the sensor signal are predicted.After processing,the signals are decomposed in time domain,frequency domain and wavelet packet respectively,and138 sensitive features of milling cutter wear condition are extracted,and then the relationship between different feature values and wear values is analyzed.In order to reduce the complexity of the prediction model,21 optimal features are selected as the input vector of the tool wear prediction model through the feature importance evaluation analysis of the random forest algorithm,and the SMOTE oversampling technology is used to balance the wear condition categories.The final feature matrix dimension is 540*21.(2)Recognition of milling cutter wear condition based on Bagging integrated learning.The relevant theory of bagging is introduced,and a representative forest random wear condition evaluation model is established on the basis of the bagging frame.Compared with various traditional machine learning algorithms,the overall performance of integrated learning algorithms such as RF,GBDT and Adaboost is better than that of single learners such as KNN and SVM,but GBDT and SVM are prone to overfitting.To this end,the idea of Bagging is introduced,SVM and GBDT algorithms are calculated in parallel to improve the generalization ability of the model,and Bagging-SVM and Bagging-GBDT milling cutter wear condition recognition models are constructed,and parameter settings and model tuning are carried out.The results show that the random selection of features during model training can effectively improve the prediction performance and generalization ability of the model.The experiment shows that the F₁ value of the Bagging-GBDT model is 0.99350,which is0.2%,8.5%and 13.2%higher than that of the Bagging-based random forest algorithm,Bagging-SVM and base learner GBDT model,respectively,which verifies Bagging-GBDT Good prediction accuracy and stability.(3)Milling cutter wear prediction based on Stacking integrated learning.This paper firstly studies the application of XGBost and Light GBM algorithms in the prediction of milling cutter wear.The comparison found that XGBoost has a better prediction effect,with MSE of 5.0253,MAE of 1.2397 and R^2 of 0.9962.In order to further improve the prediction effect of the model,a fusion model based on Stacking integration is proposed.XGBoost,Light GBM and RF are used as the primary model,and linear regression is used as the secondary model to construct the fusion model structure.The results show that the mean square error and square absolute error of the Stacking ensemble learning prediction model are 0.0026 and 0.0324,and R^2 is 0.999.Compared with the other three base learners,the Stacking ensemble learning model has better prediction performance.